dryarzeg

Now, allow me to introduce you to the ID verification. Insert your biometric passport to proceed. /j

Don't give them clues, please /s

> Try asking Chinese models about Taiwan independence or Falun Gong or the Dalai Lama or Tiananmen or the Hong Kong national security law or ASIO’s investigations into Chinese interference in Australian politics

It depends "where" you're asking. In most cases (like with DeepSeek or Z.AI models) it will gladly tell you everything (though it can hallucinate sometimes; I guess they try to filter out such data out of the training datasets) if it's not deployed on Chinese servers and you control the system prompt. So, I guess that these guardrails - probably built into the system prompt - are deployed only on China-controlled inference servers, outside of them models are pretty much talkative.

Well, at least that was my experience. Maybe yours is different for some reasons (like temperature settings or something else), I don't know.

> Is seeing people talking about the things they don’t like something that makes you unhappy? Why?

Probably (I'm just assuming) because that person observes negative/cautious/"I don't like this because X and Y and also Z"/etc sentiment too much and feels like people are only quick to notice issues while forgetting about good sides.

It's only an assumption, though.

Not that I'm too sure about how this applies (or whether it applies at all) to other people as well, but for me personally it's easier to recall information about the band if I'm being told the band's name instead of being told "well, they are known for coloured ties". So, there is certainly some kind of effect described in action.

Now, about alphabet: again, I think it's only me, but when I try to recall it backwards, I can't do that easily. I mean, I can recall it backwards, but I need more time to do that. It's harder. I'm not sure if it's because A links to B, B to C, then C to D and not backwards, or maybe just because in school you learn alphabet from A to Z and not from Z to a - so you're kind of trained to recall it A-->Z way - but it's certainly harder for me.

At the end of the day, though, I think that everyone thinks differently. Everyone is having different internal representations for concepts (such as alphabet), so it’s not surprising that this effect may work differently for different people, or not work at all.

> would definitely affect local AI

And not only. Given your example with 300W, it will affect even things like (somewhat advanced) home servers, (mostly) personal self-hosting machines and even just gaming PCs. That would affect way too much, if I understood your thought correctly.

> Android users need to switch to Graphene.

Doesn't GrapheneOS supports only Google Pixel smartphones now? For most of the users, that would mean changing their phones beforehand. And if we're talking about common people (especially not in US), it's not even everyone who can afford that. Moreover, in my opinion, by buying Google phones you're feeding Google, and I, personally, would like to avoid that.

Have you read the next paragraph?

> I suspect there are as-yet undiscovered effects which shape the next generation. Whether that be DNA methylation, gut bacteria passing from mother to child, selection of the 'correct' egg or sperm out of millions, or something new and un-discovered etc.

I can't see where it mentions "creationism".

So... is this literally a... umm, sorry, I'm just genuinely (really, no sarcasm intended) which terminology to use... finetune of DeepSeek V4-Pro or post-trained version of DeepSeek V4-Pro Base? Because I haven't fully dived into the tech report (so I may update my opinion as well as my comment), but this far the architectural solutions seem to be largely similar to DeepSeek ones.

Maybe I'm wrong, but that's just the first impression.

EDIT: I take my words back (which happens rarely) - although they do build upon DeepSeek's work, their contribution far exceeds merely post-training the base model in a different way. They did introduce something new to the architecture, though I still can't find the full tech report, with Hugging Face and GitHub links returning 404 right now.

EDIT-2: Now when I think about it, I'm not quite sure if they're going to release in the open the full report with methodology, as well as the model weights, at all.

Well, that makes some sense. My friends tried some bridges as well, but some of them didn't work from the beginning, while others got blocked (probably; the exact reason why they don't work anymore is unknown) later.

Thank you for sharing the reliable information from (well, at least somewhat) official source.

I guess that author mentions it explicitly:

> In dynamic languages like Ruby and Python tests are important for a different reason

> tests are important

> Now I can generate that same monstrosity on my own.

For what? Why would you need to?

> I've created a 1.1m+ loc and growing trading platform that automates options trades based on live ML edges from research papers.

Sounds like a lot of "if/else" /j